deepspeedai / deepspeedai/DeepSpeed

When using zero stage 3 for model training, loading custom parameters failed and the model parameter size was 0.

Open
#4,534 6 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug training
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

When I use zero stage 3, I want to load the model and then load some customized parameters. However, after loading the model, the size of the model parameters is 0. When using the safe_get_full_fp32_param method, it is displayed that the current parameters do not have _z3_optimizer. What should I do? How to solve this problem?
Thanks

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the zero stage 3 training flow that loads a model and then customized parameters. Trace the parameters involved in safe_get_full_fp32_param and the expected _z3_optimizer state; the issue is resolved when the cause of the zero parameter size is identified and a verified fix or guidance is provided.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
distributed-systems, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.